DESCRIPTION OF THE MAIN PROGRAMS

Simulations

It is worth noting that, in all subsequent programs, the synapses have been already built with a training procedure, as illustrated in the text, and the synapses so obtained are loaded from the data files “sinapsi_SET1.mat”, “sinapsi_SET2.mat” or “sinapsi_SET3.mat” for the three patterns of objects.

Figure3_alpha.m – a program that reads the patterns to be used as input and shows the temporal activity of a pyramidal neuron (in the current program, the neuron at position 10 in layer 1, see line 25) in the absence of any connection. The neuron exhibits an alpha rhythm.

Figure4_pattern.m –a program that plots the figures showing the patterns used in configurations one and two. The individual sections can be run separately.

Figure5_WMeL1 – a program that computes the behavior in the WM and L1 layers in feedback

Figure6_precession – a program that computes the recovery of a list of objects and phase precession in the modality “sequence ordering memory.”

Figure7_desynchronize_fourpatterns – a program that simulates the network functioning in the modality “semantic memory,” where different patterns can be segmented together. The patterns are chosen in line 22. In this particular configuration, the first four patterns are used.

Figure8_desynchronize_multiplepatterns – a program that simulates the network functioning in the modality “semantic memory”: only the average spike density of the different objects in layer L3 is presented for brevity. The five panels represent model response when 5, 6, 7, 8, or 9 objects are simultaneously used as input in WM per 50 ms. Each simulation is shown in a different section and can be run separately.

Figure9_pathologies – a program that simulates three examples of model behavior in pathological conditions: first section: Alzheimer; second section: schizophrenia in the sequence ordering modality; third section: schizophrenia in the semantic modality

Figure10_dreaming – a program that simulates a dreaming condition. The third sequence of objects with partial superimposition should be loaded to have interesting behavior.

Training

Training.m – This program performs a new training, divided into three steps (layer L1, L2-L3, and from L3 to L2), and saves the synapses in the arrays Wp_L1L1, A_L2L2, A_L3L3, K_L2L2, K_L3L3, and Wp_L2L3.